Streetwise Professor

The BIS has released a report titled “Macroeconomic impact assessment of OTC derivatives regulatory reforms.” It concludes that the benefits of these reforms-including Frankendodd-will greatly exceed the costs, because (a) financial costs of a financial crisis are immense, and (b) the reforms will greatly reduce the probability of a financial crisis.  Specifically:

In its report, the MAGD focuses on the effects of (i) mandatory central clearing of standardised OTC derivatives, (ii) margin requirements for non-centrally-cleared OTC derivatives and (iii) bank capital requirements for derivatives-related exposures. In its preferred scenario, the group found economic benefits worth 0.16% of GDP per year from avoiding financial crises. It also found economic costs of 0.04% of GDP per year from institutions passing on the expense of holding more capital and collateral to the broader economy. This results in net benefits of 0.12% of GDP per year. These are estimates of the long-run consequences of the reforms, which are expected to apply once they have been fully implemented and had their full economic effects.

Don’t believe it for a minute.  The methodology of the analysis is irretrievably flawed.  The net benefits are the free-est of lunches, because the report does not take into account the redistributive aspects of the collateralization and netting that results from clearing, capital, and uncleared derivatives collateral mandates.  That is, it fails to recognize that the reduction in counterparty credit losses on derivatives that result from the “reforms” are accompanied by an increase in losses suffered by somebody else.  It accounts for a transfer as a social gain.

In broad strokes, the BIS analysis goes as follows.  Financial institutions are connected to one another by OTC derivatives trades.  One way of measuring this interconnection is the CVA-the credit value adjustment on derivatives trades.  The CVA is (roughly speaking) loss given default per dollar of exposure times credit exposure times the probability of default.  If a given bank, B say, takes a hit of X to its balance sheet, due to a bad investment, or a rogue trader loss, or whatever, its probability of default goes up.  This raises the CVA.  This is immediately recognized as a mark-to-market loss, dollar for dollar, by its counterparties.  These counterparties tend to be systemically important banks, and the loss makes them more leveraged.  The greater leverage makes them riskier, and more subject to default.  This raises their probability of default, which imposes CVA losses on their counterparties, and on and on.  In this way, a shock to one bank propagates throughout the system, and feedback effects intensify the original effect.  The end result is that major financial institutions become more leveraged, and the more leveraged the system, the more prone it is to a very costly crisis.

But Frankendodd lurches to the rescue!  By collateralizing and increasing netting*, Frankendodd reforms reduce credit exposure in derivatives.  Therefore, the CVA hit from a given rise in the probability of default is smaller.  The leverage in the system doesn’t rise nearly as much when bank B takes the hit of X to its balance sheet.  So a given shock has a substantially smaller impact on systemic leverage, and hence systemic risk.  Yay!

Not so fast, people.  Notice the flaw in the logic?  Remember what started the cascading effect in the pre-Frankendodd world: an increase in the probability of default at B due to some adverse shock to its balance sheet.  Collateral and clearing mandates reduce the exposure of derivatives counterparties to this shock.  But a default doesn’t just affect derivatives counterparties: it hurts all of B’s creditors.  Roughly speaking, the total cost incurred by B’s creditors as a result of an increase in the probability of default due to the loss of X is loss given default times total liabilities times the change in the probability of default.  Given total liabilities, this does not change when derivatives are collateralized or netted more extensively.  This, in turn, means that the decline in derivatives CVA that results from more derivatives collateral and netting is matched by an increase in losses suffered by other creditors.  The “reforms” shift around the losses: they do not reduce them in aggregate.

That is, the “reforms” don’t reduce the losses that result from the balance sheet shock that raises B’s probability of default: they redistribute them.  Note, moreover, that many of the same big financial institutions that benefit from the decline in derivatives CVA are hurt by the loss on B’s other liabilities, because these institutions are exposed to one another through a variety of claims, not just derivatives.  Moreover, some of the others (non-bank creditors) that suffer from the redistribution may be systemically important too: money market funds that invest in the short term debt of financial institutions (including B) are one example.  Repo is another example.  Thus, the overall effect of Frankendodd and EMIR on systemic risk is quite equivocal.  It shifts around losses, and there is no guarantee that the shift in losses improves the stability of the financial system.

But the BIS study does not treat the redistributive effects of the derivatives reforms.   It treats them has a net gain.  This is fundamentally, basically, and irredeemably wrong.  Wrong, wrong, wrong.  When evaluating systemic risk, it ignores the systemic redistribution of losses.  As a result, it overestimates the gains of the regulatory changes.

To a first approximation, given the redistributive nature of their effects, the benefits of the reforms is zero.  The BIS recognizes that they involve costs.  Meaning that to a first approximation, Frankendodd and EMIR reduce wealth, rather than increase it.

But that’s just a first approximation.  To understand fully the effects, you’d have to know how losses are redistributed, and the systemic importance of those who suffer bigger losses and of those who realize smaller ones.  You’d also have to understand how financial contracts will change in response to the new set of creditor priorities inherent in the Frankendodd and EMIR rules.   Financial contracts-capital structures, if you will-will change in response to the new rules.  Claims will be repriced.  There is going to be a new, different equilibrium structure of financial contracts.  Maybe this new structure is less fragile.  Maybe it is more so.  I don’t know: the system is so complex that it will respond to this big shock in surprising ways.  All I do know is that this where you need to look to figure out the effects of the so-called reforms on systemic risk, and this ain’t where the BIS looks.  It counts benefits conferred on some subset of claimants, while ignoring the costs imposed on others (that are approximately equal in magnitude).

In other words, the BIS is hawking a free lunch, and as Friedman said, there ain’t no such thing.

The BIS study also has a particular model of systemic risk: the counterparty credit contagion model.  B takes a hit, that spreads to C, D, and E, which then spreads to F, G, and H, and on and on, until most everyone drops dead like Aztecs fell to the smallpox.  This is a common way of formalizing systemic risk, but reminds me of the story of the drunk looking for his keys under the lamppost because the light is better there, not because he lost them there.  If you look at the history of financial crises, they almost never look like the systemic crises in these contagion models.  An idiosyncratic shock at one institution doesn’t bring a house of cards crashing down.  The collapses of individual institutions (Bear, Lehman) are symptoms of a deeper-and systemic-rot.

In sum, the BIS analysis is fundamentally flawed, and hence gives a wildly misleading estimate of the social benefits of the clearing and collateral mandates embedded in Frankendodd and EMIR and regulations adopted by other G20 nations.  It is a sad, sad example of a reputable institution falling prey to free lunch fallacies.

It is particularly sad because it’s not as if this hasn’t been pointed out before.  I raised the problem five years ago.  So you may say-well, who are you anyways?  Well, Harvard’s Mark Roe, one of the most accomplished bankruptcy scholars in the legal academy, has made the same point (and graciously cited my work).  And, truth be told, it is a staple of the bankruptcy literature.  And it’s also sad because I hear such “logic” over and over from regulators (e.g., scholars and policymakers at the Fed).  They presume to identify systemic benefits based on analyses that look at only a portion of the system.

And speaking of sad.  Could the BIS please spring for a typesetting program like LaTeX that prints readable equations?  Please?

* It is by no means clear that moving more trades to clearing will reduce credit exposures in derivatives via netting, especially given the fragmentation of CCPs by product and jurisdiction.  CCPs permit multilateral netting, but (a) bilateral trades can be compressed multilaterally, and (b) cross-product netting of exposures within dealer books can be larger than multilateral netting in single-category CCPs.  But that’s really a secondary issue.  The most important thing to remember is that the primary effect of netting is to redistribute default losses from derivatives counterparties to non-derivatives creditors.